1 Introduction

2 Methods

The ACS-NSQIP database was queried for patients who had undergone any of the following procedures – total hip arthroplasty (THA), total knee arthroplasty (TKA), anterior cruciate ligament (ACL) repair/reconstruction, microdiscectomy, anterior cervical discectomy and fusions (ACDF), one-level lumbar fusion, distal radius fracture fixation, or olecranon fracture fixation – from 2005 to 2016.^4,5 These patients were isolated based on their primary Current Procedure Terminology (CPT) codes, outlined in Appendix A. Non-elective surgeries, emergent surgeries, those with invalid operative times, and those with known resident-physician involvement were excluded. 536,274 patients were ultimately included for analysis.

In order to calculate the time-cost difference, as defined by, “Time-Cost Difference = [(Average time it takes to complete procedure as an inpatient) – (Average time it takes to complete procedure as an outpatient)] x $37/minute,” patients within each procedural group were stratified based on whether they had undergone inpatient or outpatient surgeries.³ The sample of 536,274 patients was analyzed as a whole, as well as within each respective procedural group, with Independent-Samples t-tests, Pearson's chi-squared tests, and Fischer's exact tests (expected values < 5). T-tests were utilized to identify differences in mean operative times between the inpatient and outpatient patient cohorts, while the chi-squared tests/Fischer's exact tests were utilized to calculate differences in demographics & preoperative comorbidities (i.e. race, sex, diabetes mellitus, hypertension, etc.).

All statistically significant differences in patient demographics and preoperative comorbidities were controlled for through propensity-score matching in a randomized 1:1 manner without replacement, with the exception of olecranon fracture repair; these patients were matched in a 4:1 manner due to the smaller sample size. The caliber was set at 0.2, and the propensity-generated cohort of inpatient procedures was subsequently compared to those of the outpatient cohort for any differences. Univariate analyses were utilized in assessing differences between the two groups. All statistical findings with p-values less than or equal to 0.05 were considered significant in this analysis. Specific demographic factors and comorbidities with statistically significant differences in rates between inpatient and outpatient cohorts were controlled for by generating multivariate logistic regression models. Propensity-score matching was completed using R© (2017) and IBM ® SPSS® Statistics Version 25 software. (IBM Corporation, Armonk, NY).

Multivariate logistic regression models were generated to assess outpatient surgeries as an independent risk factor for adverse outcomes (i.e. complications, reoperation, extended length of stay ≥5 days (LOS), readmission, & mortality) in comparison to inpatient surgeries. This study defined “minor complications” as transfusions, pneumonia, wound dehiscence, urinary tract infection, renal insufficiency, and superficial surgical site infection (SSI). “Major complications” were defined as including deep incisional SSI, organ space SSI, unplanned intubation, pulmonary embolism, failure to wean (on ventilator > 48 h), renal failure, stroke, cardiac arrest requiring cardiopulmonary resuscitation, myocardial infarction, deep venous thromboembolism, systemic sepsis, septic shock, reoperation, death, extended LOS, and readmission. Multivariate logistic regressions assessing outpatient surgery as a risk factor for adverse outcomes were not generated for THA, lumbar fusion, and olecranon fracture fixation, since both procedures had patients without statistically significant differences in demographic factors or preoperative comorbidities when stratified by inpatient/outpatient status. In these instances, chi-squared and Fischer's exact tests were utilized to assess differences in complication rates.

Multivariate logistic regression models were generated to assess longer operative times as an independent risk factor for adverse events (i.e. complications, reoperation, extended LOS, readmission, & mortality) for the various procedures. Longer operative times were defined as those in the 75th percentile for each respective procedure length. These multivariate analyses controlled for significantly different demographics and preoperative comorbidities when stratified into those with longer operative times and those without. All statistical analyses were performed using the IBM® SPSS® Statistics Version 25 software. (IBM Corporation, Armonk, NY).

3 Results

Inpatient orthopaedic procedures generally demonstrated significantly longer aggregate operative times (x̄ = 134.59 min) when compared to outpatient settings (x̄ = 88.21 min; p < 0.001). Performing the examined procedures in the outpatient setting demonstrated a time-cost savings of $1716.06 (Table 1). Multivariate regression analyses demonstrated increased rates of major and minor complications, reoperation, extended LOS, and unplanned readmission in the inpatient setting when compared to the outpatient setting (p < 0.001) (Table 2). Longer operative times also increased the risk for the same aforementioned postoperative complications when taking the orthopaedic procedures in aggregate (p ≤ 0.001) (Table 2).

Of the procedures analyzed, TKA was one of two procedures that failed to demonstrate significant differences in mean operative times between inpatient and outpatient settings (93.94 min vs. 93.13 min; p = 0.506) (Appendix B). Outpatient TKAs compared to inpatient TKAs demonstrated increased risk of extended LOS (p = 0.042) (Appendix B), but multivariate analyses failed to demonstrate outpatient setting as a risk factor for major complications, minor complications, reoperation, readmission, and death (Appendix B). Longer operative times in TKA also failed to demonstrate an increased risk for major complications, minor complications, reoperation, extended LOS, and readmission (Appendix B).

THA also failed to demonstrate a significant difference in mean operative times between inpatient and outpatient care settings (92.16 min vs 90.89 min; p = 0.426) (Appendix C). THAs performed in the outpatient setting experienced higher rates of major complications (p = 0.021) and extended LOS (p = 0.018) (Appendix C). However, inpatient THAs experienced higher rates of minor complications (p = 0.014) (Appendix C). With regard to longer operative times in THA, longer operative times increased the risk of any complication (p < 0.001), major complications (p = 0.003), minor complications (p < 0.001), and extended LOS (p < 0.001) (Appendix C).

Inpatient ACL repairs and reconstructions demonstrated longer mean operative times compared to outpatient repairs (133.19 min vs 102.04 min; p < 0.001) for a time-cost difference of $1152.55 (Appendix D). Inpatient ACL repairs/reconstructions demonstrated increased risk for any complication (p = 0.014) (Appendix D). Longer operative times for ACL repairs demonstrated increased risk for any complication (p < 0.001), major complications (p = 0.001), reoperation (p = 0.003), and extended LOS (p = 0.005) (Appendix D).

Inpatient microdiscectomy procedures demonstrated longer mean operative times compared to outpatient settings (139.6 min vs 81.49; p < 0.001) for a time-cost difference of $2109 (Appendix E). Aside from mortality (p = 0.101), inpatient settings for microdiscectomies increased the risk for the remaining complications included in this analysis (p < 0.001) (Appendix E). Microdiscectomies that had longer operative times increased the risk for all complications (p < 0.001; death: p = 0.004) (Appendix E).

ACDF demonstrated a significant difference in mean operative times between inpatient and outpatient surgical settings (130.56 min vs 101.16 min; p < 0.001) yielding a time-cost difference of $1087.80 (Appendix F). ACDFs done in inpatient settings experienced increased risk for all complications (p < 0.001; death: p = 0.044) (Appendix F). Apart from death, longer operative times in ACDF increased the risk of the remaining complications in the postoperative period (p < 0.010; death: p = 0.240) (Appendix F).

One-level lumbar fusion procedures performed in inpatient settings demonstrated significantly longer operative times when compared to outpatient settings (206.59 min vs 138.55 min; p < 0.001) for a time-cost difference of $2517.48 (Appendix G). Lumbar fusion procedures done in the inpatient setting experienced significantly higher rates of any complication (p < 0.001), major complication (p < 0.001), minor complication (p < 0.001), and extended LOS (p < 0.001) when compared to the outpatient cohort (Appendix G). Longer operative times in one level lumbar fusions demonstrated significantly increased risk for any complication (p < 0.001), major complication (p < 0.001), minor complication (p < 0.001), reoperation (p = 0.021), and extended LOS (p < 0.001) (Appendix G).

With a time-cost difference of $537.98, distal radius procedures done in the inpatient setting were significantly longer than done in the outpatient setting (89.59 min vs 75.05 min; p < 0.001) (Appendix H). Inpatient distal radius procedures were associated with increased risks for any complication (p < 0.001), major complication (p < 0.001), and extended LOS (p < 0.001) (Appendix H). Risks of death and readmission were not increased with longer operative times. Distal radius procedures with longer operative times demonstrated significantly increased risks for the same complications (p < 0.005) (Appendix H).

Inpatient olecranon fractures demonstrated significant differences in mean operative time between the inpatient and outpatient care settings (94.43 min vs 81.62 min; p < 0.001) with a time-cost difference of $473.97 (Appendix I). Olecranon fractures treated in the inpatient setting experienced significantly higher rates of any complication (p < 0.001), major complication (p < 0.001), and extended LOS (≥5 days) (p < 0.001) when compared to the outpatient cohort (Appendix I). With the exception of extended LOS (p = 0.004), olecranon fractures with longer operative times failed to demonstrate any associated increase in risk of any other complications (Appendix I).

4 Discussion

The use of large databases in orthopaedic research is increasing.^6–9 The large number of patients (536,274) included in this analysis is strength of this study, adding certainty behind the statistical calculations. Also, because the ACS-NSQIP is a national database, the findings in this study are more likely to be applicable to US providers regardless of geographical setting. As opposed to other databases which are more focused on billing purposes or large collections of insurance claims data, the NSQIP was specifically collected to improve surgical quality.¹⁰ The data entered into NSQIP are by medically trained professionals as opposed to other databases. The data is also reviewed by the NSQIP Data Definition Committee and interrater reliability disagreement is less than 1.8%.¹⁰

The ideal study to compare differences in operative times and outcomes for procedures performed in an outpatient or inpatient setting would be a randomized controlled study (RCT). However, in lieu of designing such an involved RCT, this database review, with its large volume of patients matched for medical co-morbidities and demographics, presents a reasonable alternative from which to draw conclusions.

Orthopaedic research utilizing large national databases come with inherent risks. Recent research has demonstrated that the same clinical question, when posed to different databases, may result in differing results.^6,11 Research utilizing databases is also inherently retrospective in nature. The NSQIP only captures data within the 30 days immediately postoperatively, limiting the inclusion of adverse events occurring outside this window. Databases allow for large n-values which may determine small but statistically significant differences. However, the reader must make the distinction whether these statistically significant differences represent clinically relevant findings. The NSQIP was designed for surgeon use and its outcomes are not specific to orthopaedic surgery. Although adverse events are critical in determining patient outcome, satisfaction, range of motion, and pain are additional important factors in orthopaedics and are not included in the NSQIP.⁷

All procedures in our sample indicated faster operative times in the outpatient setting except for total knee and hip replacements. This suggests that there are significant potential cost savings to be realized if policymakers and providers place a greater emphasis on performing elective orthopaedic procedures in an outpatient setting. On average, the cost savings are $1716.06 per procedure with the greatest statistically significant differential in one-level lumbar fusions ($2517.48) and the smallest statistically significant differential in olecranon fracture fixation ($473.93).

The lack of statistically significant operative time differences between inpatient and outpatient total joint replacements can perhaps be explained by the unique payment structure to providers for these procedures. Health insurers, including Medicare, have largely transitioned from fee-for-service reimbursement to global payments intended to cover the entire episode of care over a 90-day period.¹² Providers make a profit if the total cost of care is less than the payment received from insurers and suffer a loss if the total cost of care is more expensive. With capitated payments, providers are incentivized to employ cost saving measures. This incentive would naturally drive inpatient providers to streamline operations such that the throughput of their total joint patients matches that of outpatient centers. This is reflected in operative times that are statistically equivocal. For procedures that are not covered under a global payments program, no such incentive for improving inpatient efficiency exists and thus the operative time and cost differences remain.

The analysis of outpatient setting as an independent risk factor for complications, re-operations, extended LOS, and readmission show that, in aggregate, procedures performed in an outpatient setting have better outcomes (e.g. lower odds ratios) after controlling for patient co-morbidities and demographics. With the exception of ACDFs, which did show statistically significantly lower odds, procedures in an outpatient setting did not affect the odds ratio for mortality. Looking at this analysis for each procedure individually presents interesting findings. Paradoxically, with TKA we find that there are increased odds of having an extended LOS when the procedure is performed in an outpatient setting. This finding is difficult to explain, and perhaps is anomalously statistically significant given the low R-squared value of the regression model. Similarly, with outpatient THAs we find that there are increased rates of having an extended LOS and a major complication but decreased rates of having a minor complication in the outpatient setting. An alternative explanation may be that when complications in outpatient total joints do occur, they are serious enough to require a longer LOS (e.g. infection requiring IV antibiotics). Future work via a RCT may prove beneficial in further delineating the impact of an outpatient setting on postoperative complications for THA and TKA.

Outpatient ACLs, distal radius fracture fixations, and olecranon fracture fixations have lower odds of a complication occurring compared to their inpatient counterparts. For olecranon and distal radius fractures, there are also lower odds of an extended LOS when fixation is performed on an outpatient basis. All spine procedures (i.e. microdiscectomies, ACDFs, and one-level lumbar fusions) have lower odds of complications, re-operations, and extended LOS in the outpatient setting. ACDFs and microdiscectomies also have statistically significantly lower odds of readmission, though the same is not true for one-level lumbar fusion.

The analysis of operative time as an independent risk factor for complications, re-operations, extended lengths of stay, readmission, and death show that, in aggregate, longer operative times result in worse outcomes (e.g. higher odds ratios) after controlling for patient co-morbidities and demographics. However, with TKAs in particular, longer operative times did not result in any statistically significant change in odds ratios. For olecranon fractures, longer operative times were associated with higher odds of an extended LOS but not for any of the other outcome variables. This suggests that longer operative times are less of a detriment in TKAs and olecranon fractures. The opposite appears true for spine procedures. Longer operative times in microdiscectomies have higher odds ratios of complications, re-operations, re-admissions, and even death. Likewise, there are higher odds ratios of a negative outcome for longer operative times for ACDFs (excluding mortality) and one-level fusions (excluding mortality and readmission). For THA and ACL repairs, longer operative times increased the odds ratio of developing a complication or having an extended LOS. For THA, longer operative times were also associated with increased odds of reoperation.

5 Conclusion and implications

Hospital care in the US is becoming increasingly more expensive, and without changes to the funding structure, Medicare will be unable to fully cover inpatient costs as soon as 2026. All stakeholders – insurers, providers, and patients – must search for lower-cost care strategies to build a more sustainable health system. One such strategy may be to encourage more elective orthopaedic procedures to be done on an outpatient basis, or alternatively to incentivize inpatient providers to be as streamlined and efficient as their outpatient counterparts. Our study shows that doing so would result in cost savings of $1716.06 per procedure due to shorter operative times.

Future research might focus on subdividing the outpatient surgery setting into ambulatory surgery centers (ASCs) versus hospital outpatient departments (HOPDs) and to perform a similar comparative analysis of operative times, costs, and quality outcome measures. Unfortunately, such a subdivision was not possible using the ACS-NSQIP database, but the implications from such an analysis could be far-reaching. If operative times in ASCs are shorter, then policymakers should potentially push for ASCs as the ideal operative setting in order to save the health system costs. Such a policy focus could represent a win for all stakeholders. Insurers such as Medicaid and Medicare would save money. Given the high rate of physician-ownership in ASCs, surgeons would assume more responsibility for patient care and gain financially from earning both facility and professional fees.¹³ Most importantly, patients would benefit from having shorter procedures with better outcomes. Global payments may represent one way of incentivizing inpatient providers to perform more effectively as evidenced by the lack of a statistically significant difference in operative times between total joint replacements in the outpatient versus inpatient setting.

This paper suggests that a shift to more elective outpatient procedures would not compromise quality of care; in fact, it may potentially improve quality, given the lower odds ratios of developing adverse patient outcomes in the outpatient setting with shorter operative times.

IRB approval statement

This study was exempt from Institutional Review Board approval as the ACS-NSQIP database is a nationally publicly available database of de-identified patient information.

Funding

No funding was received for this work.